Eddie On Everything

The copy command “cp -al” found on all versions of Unix/Linux creates what’s called a “hard link” to a file. The nice thing about this command is that it doesn’t create an actual copy of the file on disk – instead, it creates a “link” or pointer to the file data on the disk. Basically a “snapshot” of that directory in-time. The net result is that you can have 10 “copies’ of a 10G file that only take up a total of 10G.

This nifty behavior makes cp -al, when combined with rsync, ideal for backup systems. One can use the cp -al command to take a “snapshot” of a given directory tree at a given time, at the expense of very little additional disk space. I use this script in concert with my rsync_backup.pl script to keep 21 days of “snapshot” backups of each of my machines.

#!/usr/bin/perl
use POSIX;

# Rotates backup directories w/ cp -al (hardlinks)
# Deletes directories older than $KEEP_DAYS
# Runs each night ahead of backup process
# (c) 2009 eddie@eddieoneverything.com

$KEEP_DAYS=21;
$LOGFILE = "/var/log/rotate_backups";

@BACKUP_DIRS=(
   '/mnt/backup/hansel',
   '/mnt/backup/tiger',
   '/mnt/backup/june'
);

$ts = get_timestamp();

open hLOG, ">>$LOGFILE";
print hLOG "=" x 80, "\n";
print hLOG "Run START at " . `date` . "\n";
print hLOG "=" x 80, "\n";

## Do the rotation
print hLOG "Do today's rotation\n";
foreach $dir ( @BACKUP_DIRS ){
   print hLOG "\t" , `date`;
   $newfn = $dir ."_" .  $ts;
   $cmd = "cp -al $dir $newfn";
   print hLOG "Execute Command: $cmd\n";
   `$cmd`;
}

## Delete old directories
print hLOG "Delete Old Directories\n";
foreach $dir ( @BACKUP_DIRS ){
   $dir=~m/^(.+)\/(.+?)$/;
   $base= $1;
   $stub=$2;
   #print "dir is $dir\nBASE: $base\nSTUB:$stub\n";

   opendir hDIR, "$base" or die "can't open directory $base";
   @dirlist=grep { /^$stub\_/ && !/^$stub$/  } readdir(hDIR);
   closedir hDIR;

   foreach $d (@dirlist){
      print hLOG "\t" , `date`;
      $d=~/$stub\_([0-9]+)_([0-9]+)_([0-9]+)_.+$/;
      $year= $1;
      $month = $2;
      $day = $3;

      if (dirIsOlder($year, $month, $day)){
         #print "$d\n";
         $remove_dir= $base . '/' . $d;
         $cmd = "rm -Rf \"$remove_dir\"";
         print hLOG "Execute command: $cmd\n";
         `$cmd`;
      }else{
         print hLOG "Keep $d\n";
      }
   }
}
print hLOG "=" x 80, "\n";
print hLOG "Run END at " . `date` . "\n";
print hLOG "=" x 80, "\n";
close hLOG;

# ------------- Subroutines & functions -----------------

sub get_timestamp {
   ($sec,$min,$hour,$mday,$mon,$year,$wday,$yday,$isdst) = localtime(time);
   $mon+=1;

   if ($mon < 10) { $mon = "0$mon"; }
   if ($mday < 10) { $mday = "0$mday"; }
   if ($hour < 10) { $hour = "0$hour"; }
   if ($min < 10) { $min = "0$min"; }
   if ($sec < 10) { $sec = "0$sec"; }
   $year=$year+1900;

   return $year . '_' . $mon . '_' . $mday . '__' . $hour . '_' . $min . '_' . $sec;
}

sub dirIsOlder{
   ($fyear, $fmonth, $fday) = @_;
   #print "Check $year - $month - $day \n";

   $now  = mktime(localtime());
   $then =  mktime (0, 0 , 0, $fday-1, $fmonth-1, $fyear-1900 , 0, 0);

   $diff_sec = $now - $then;
   $days_since = $diff_sec / 24 / 60 / 60;
   #print "n: $now. t: $then. Diff ($diff_sec) = $days_since\n";
   #print "$days_since $fyear-$fmonth-$fday\n";

   ## Subtract days from today

   if ($days_since > $KEEP_DAYS){
      return 1;
   }else{
      return 0;
   }
}